In a PCM recorder for recording/reproducing a PCM audio signal, a few words each comprising 16-bit data are given as one block. A data error detection/correction code such as a CRC and a parity word is added to each block for detection and correction of a read error bit in the reproduction mode. Normal random and burst errors can be corrected to proper data by using an error detection correction code. However, when a great number of errors occur and cannot be corrected, the error positions are detected, and error correction such as interpolation processing and the holding of the previous value is performed. The interpolation is performed by using two data values adjacent to error data.
On the other hand, error data which is neither detected as an error nor corrected is present. Assume a PCM signal is recorded on a recording medium such as a magnetic tape and the recorded PCM signal is erased to record another PCM signal. When the previous PCM signal is not completely erased due to the presence of dust or the like and the partially left previous PCM signal is reproduced, it cannot be detected as an error by an error detection/correction circuit (check circuit). As a result, error correction or compensation will not be performed, and pulse noise (abnormal sound) such as click noise inherent to PCM recording is generated. In a VTR system using a magnetic tape and rotary heads, when the recorded tape is reproduced at a speed (e.g., slow play or fast play) different from that in the recording mode, or mistracking occurs, the rotary head scans across a plurality of tracks. In this case, an error is neither detected nor corrected. Therefore, an abnormal sound caused by pulse noise is produced.